JP5644087B2 - Component highlighting apparatus, program, and method - Google Patents

Description

  More particularly, the present invention relates to a component highlighting device.

Patent documents must be read by non-patent experts for the purpose of inventors to avoid infringement of other companies' patents.
However, claims and the like described in claims of patent documents are generally written in words or combinations of words, and thus are often written in abstract words with a specific meaning. It is difficult for an inventor who is unfamiliar with patent documents to grasp the points of the invention described in the claims.

  In addition, in the specification of a patent, in addition to the constituent elements that are the points in the invention, the entire surrounding constituent elements are also described, so when reading the specification, the constituent elements that are the points of the patent Need to read while looking for an explanation about.

On the other hand, in a document describing the effects of the invention, research results have been reported that it is possible to extract problem expressions and effect expressions by using an expression pattern such as “can do it ... Yes. (For example, Non-Patent Document 1)
In addition, in a document describing the effect of the invention, for example, using an expression pattern such as “depending on ...”, a relationship (causal relationship) between the description of the means and the description of the effect can be extracted. Research results have also been reported. (For example, Non-Patent Document 2)

JP 2002-63192 A

Y. Sakachi, Hiroshi Nonaka, Hiroyuki Sakai, Shigeru Masuyama, Extraction of problem / effect expression pairs from patent documents using bootstrap method, IPSJ research report, vol. 2009-NL-192, no. 14, pp. 85-92, 2009. Daisuke Ishikawa, Hidehiro Ishizuka, Norihiko Uda, Joe Fujiwara, Extraction and integration of causal relationships in patent literature: Overview and subsequent development, Journal of Information Knowledge Society, Vol. 15, pp. 98-106, 2005.

However, conventionally, there has been no known technique that can present a constituent element that is a point in a patent document.
Therefore, an object of the present invention is to identify and present a constituent element as a point in the claims so that the explanation of the constituent element as the point can be read with emphasis.

  A component highlighting device according to one aspect of the present invention is a device for supporting understanding of a patent document, and is a character string described before or after a component name corresponding to a component constituting a claim. A component name extraction unit that extracts each component name from the text data of the claims in the patent document extracted from the patent document database using the component expression pattern dictionary in which the pattern is recorded, and each component name and component Based on the character string pattern recorded in the expression pattern dictionary, from the text data of the claim, the text data of the descriptive text explaining the constituent element corresponding to each constituent element name is extracted. The effect of recording the character string pattern that links the component's explanation extractor, the character string part describing the effect, and the character string part describing the reason for obtaining the effect. A reason extraction unit that can obtain the effect of extracting the text data describing the reason for obtaining the effect from the text data describing the effect of the invention in the patent document using the reason expression pattern dictionary , A similarity calculation unit for calculating the similarity between the text data of the explanatory text of each component and the text data describing the reason why the effect is obtained, and a configuration corresponding to the text data of the explanatory text with the highest similarity A point component specifying unit that specifies an element as a component that becomes a point, and an output unit that outputs the component that becomes a point are included.

  In the document describing the effect of the invention in the patent document, it is possible to identify a component that has a strong relationship with the reason why the effect is obtained and present it to the user as a component that becomes a point.

It is a block diagram of embodiment. FIG. 3 is a diagram (part 1: Japanese Patent Laid-Open No. 2002-278562) showing an example of a patent document. It is operation | movement explanatory drawing of embodiment (the 1: process which extracts a component element name from a claim). It is operation | movement explanatory drawing (2): The process which extracts the sentence which has demonstrated the component from the claim. It is operation | movement explanatory drawing (3: The process which extracts the part which has written the reason for obtaining an effect) of embodiment. It is operation | movement explanatory drawing (No. 4: The process which specifies the component in search of the description of the component most similar to the reason from which an effect is acquired). It is a figure which shows the example of a patent document (the 2nd: patent 2754166 gazette). It is operation | movement explanatory drawing (5: the process which extracts a component from a claim) of embodiment. It is operation | movement explanatory drawing (6: the process which extracts the sentence which has demonstrated the component from the claim) of embodiment. It is operation | movement explanatory drawing (7: The process which determines whether there is sufficient information in the description extracted from the claim) of embodiment. It is operation | movement explanatory drawing (8: The process which extracts the sentence which has demonstrated the component from the Example) of embodiment. It is operation | movement explanatory drawing (9: The process which extracts the part which has written the reason for obtaining an effect) of embodiment. It is operation | movement explanatory drawing of embodiment (the 10: the process which calculates | requires description of the component most similar to the reason with which an effect is acquired, and specifies a component). It is a flowchart of an embodiment. FIG. 15 is a detailed flowchart of step S1406 in FIG. 14 (processing for determining whether there is a sufficient amount of information in the extracted explanatory text). FIG. 15 is a detailed flowchart of step S1409 in FIG. 14 (processing for extracting a reason for obtaining an effect). FIG. 15 is a detailed flowchart of step S1410 in FIG. 14 (processing for calculating the similarity between the reason why the effect is obtained and the description of each component). It is a figure which shows the example of a patent database. It is a figure which shows the example of an analyzed document table. It is a figure which shows the example of a component element expression pattern dictionary. It is a figure which shows the example of the explanatory expression pattern dictionary in an Example. It is a figure which shows the example of a component table. It is a figure which shows the example of the reason expression pattern dictionary from which an effect is acquired. It is a figure which shows the example of a morpheme weight table. It is a figure which shows the example of a similarity table. It is a figure which shows the example of a display of a display part. It is an operation | movement flowchart which shows the dependency analysis used for each component extraction process with a modifier. It is explanatory drawing of a dependency analysis process. It is a detailed flowchart which shows the specific operation | movement of a pattern matching process. It is a figure which shows the example of the hardware which can implement | achieve the system of embodiment.

Hereinafter, embodiments for carrying out the present invention will be described in detail with reference to the drawings.
FIG. 1 is a block diagram of an embodiment of a component highlight device. The system of the present embodiment includes a patent document search unit 101, a patent document structure analysis unit 102, a component name extraction unit 103, a component description extraction unit 104 from the claims, an information amount determination unit 105, and an explanation from the examples. An extraction unit 106, a reason extraction unit 107 for obtaining an effect, a similarity calculation unit 108, a point component specifying unit 109, and a display unit 110 are provided. In addition, the system according to the present embodiment includes a patent database 111, a component expression pattern dictionary 112, a component table 113, an analyzed document table 114, an explanation expression pattern dictionary 115 in the embodiment, and a reason expression pattern dictionary 116 for obtaining the effect. , And a similarity table 117.

The basic operation of the present embodiment having the above-described configuration will be described below.
As a clue for identifying the constituent elements that are the points in the claims, description of documents describing the effects of the invention, such as the item “Effects of the Invention” in the specification, is useful. Such a document is considered to contain the most characteristic effects of the invention, so that the constituent elements of the invention can be extracted from the description of such documents and associated with the constituent elements in the claims. If it is possible, the component which becomes a point can be specified.

Therefore, in the present embodiment, it is realized to identify and present a component that acts strongly on the reason for obtaining the effect in the document describing the effect of the invention in the patent document.
However, in a document describing the effect of the invention, a constituent element as a point may be written in place of a constituent element name instead of a description of its operation.

  For example, from the description column of “Effect of the invention” shown in FIG. 2B of the patent publication gazette, the expression pattern “can do so ... can be done” is extracted, so The underlined part of the first half corresponding to “So” is a description of the reason why the effect is obtained, and it can be recognized that the underlined part of the latter half corresponding to “... As a result, the description in the underlined part of the first half “The broadcasted song can be specified and downloaded by the time of listening and the broadcasting company, broadcasting channel, or store code”, It can be determined that the relationship with the element is strong. However, in the description of the underlined part in the first half, for example, any of the components “mobile phone”, “download server”, and “broadcast display means” described in “Claim 1” in FIG. Is not listed. Therefore, it is not possible to directly specify the constituent element that is the point in “Claim 1” from the description of the underlined portion in the first half.

  Therefore, in this embodiment, the degree of similarity between the explanatory text of each component of the invention and the document of the reason why the effect of the invention is obtained in the patent document by computer processing based on the following first procedures 1 to 4 And the most similar component is identified as the component that becomes the point. Hereinafter, these procedures will be described with reference to the example of the patent document of FIG. 2 and the operation explanatory diagrams of FIGS.

First procedure 1 : From the text data of the claims in the patent document extracted from the patent document database 111 (FIG. 1), to the expression pattern recorded in the component expression pattern dictionary 112 (FIG. 1 and FIG. 20 described later). Based on this, the component name is extracted. The text data of each component name obtained as a result is registered in the component table 113 (FIG. 1 and FIG. 22 described later).

  For example, with respect to “Claim 1” in FIG. 2A, the component names “mobile phone”, “download server”, and “broadcast display means” are extracted as indicated by 301 to 302 in FIG. 113 is registered. At this time, for example, each component is identified by using the expression pattern “having to, and to” recorded in the component expression pattern dictionary 112 for extracting the component shown in FIG. The

First procedure 2 : Extract text data describing the component corresponding to each component name extracted in the first procedure 1 from the text data of the claims. The text data of the explanatory text of each component obtained as a result is registered in the component table 113 (FIG. 1 and FIG. 22 described later) in association with each component name.

  For example, with respect to “Claim 1” in FIG. 2A, text data describing the component name “mobile phone” is extracted as indicated by 401 to 402 in FIG. 4, and the component name “mobile phone” is extracted. Is registered in the component table 113 as a description of the component. The same applies to the other component name “download server” and “broadcast display means”.

First procedure 3 : Extract text data describing the reason why the effect is obtained from the text data describing the action and effect of the invention in the patent document extracted in the first procedure 1. At this time, by referring to the expression pattern recorded in the reason expression pattern dictionary 116 (FIG. 1 and FIG. 23 described later) from which the effect is obtained, the part of the reason for obtaining the effect is extracted.

  For example, as shown by 501 to 502 in FIG. 5 for “effect of the invention” in FIG. 2B, text data describing the reason for obtaining the effect is extracted. At this time, for example, using the expression pattern “NO” recorded in the reason expression pattern dictionary 116 for obtaining the effect shown in FIG. 23, the text data of the part before the expression pattern (the underlined portion 502 in FIG. 5) is obtained. It is extracted as a part where the reason why the effect is obtained is described.

First procedure 4 : Of the text data of the explanatory text of each component registered in the component table 113 in the first procedure 2, the most similar to the text data of the reason for obtaining the effect extracted in the first procedure 3 Calculate what to do. And the component corresponding to the text data of the calculated explanatory note is specified as a component which becomes a point.

The most similar is calculated as follows.
For example, with respect to “Claim 1” in FIG. 2A and “Effect of the invention” in FIG. 2B, as shown as 602-1, 602-2, and 602-3 in FIG. The text data of the noun is extracted by the morphological analysis from the text data of the description of each component “mobile phone”, “download server”, and “broadcast display means” extracted from claim 1 in the procedure 2. The text data of these nouns can be configured to be morphologically analyzed and registered in the component element table 113 in advance corresponding to each explanatory sentence as shown in FIG. On the other hand, as shown as 601 in FIG. 6, noun text data is extracted by morphological analysis from the text data of the reason why the effect extracted in the first procedure 3 is obtained. The number of nouns included in common with the text data 601 of the reason for obtaining the effect is similar for each of the text data 602-1, 602-2, and 602-3 of the explanatory text of each component in the claims. Counted as degrees. At this time, it can be configured such that, as the frequency of appearance of each noun in the specification of the patent document is lower, the noun is more significant (higher uniqueness) and is weighted higher in the similarity count. As a result, the text data 602-1 having the largest number of significant nouns commonly included in the explanation data of the constituent elements in the claims and the text data of the reason for obtaining the effect is the text data of the constituent elements having the highest similarity. Extracted as Then, the component name “mobile phone” corresponding to the text data 602-1 is extracted from the component table 113 (FIG. 1 and FIG. 22 to be described later) and specified as the component to be a point.

  In the document describing the effect of the invention in the patent document by the processing of the first procedure 1 to 4 described above, a component that acts strongly with the reason for obtaining the effect is identified and becomes a point. It can be presented to the user as a component.

  In this case, in the document describing the effect of the invention, even if the component is written in place of the component name instead of the description of the operation, the description of the component in the claim The similarity between the text data of the text and the text data of the reason for obtaining the effect is determined. Thereby, it becomes possible to pinpoint the component which becomes a point exactly.

  In the processing of the first procedure 1 to 4 described above, the description of the constituent elements described in the claims alone does not contain sufficient information for determining the similarity in the first procedure 4 There is. For example, with respect to the component “voice converting means” in “Claim 1” of the patent document shown in FIG. 7A, “digital audio read from the reading means (6) is subjected to digital-analog conversion and analog. Only the voice conversion means (8) for outputting voice is described. This explanation contains almost no specific content. For this reason, the similarity in terms of the number of significant nouns between the reason why the effect in the “effect of the invention” shown in FIG. 7B is obtained (the underlined part in the first half in FIG. 7B). The degree cannot be determined accurately.

  Therefore, in the present embodiment, the above-described first procedures 1 to 4 are improved, and computer processing based on the following second procedures 1 to 7 is executed.

Second procedure 1 : from the text data of the claims in the patent document extracted from the patent document database 111 (FIG. 1) to the expression pattern recorded in the component expression pattern dictionary 112 (FIG. 1 and FIG. 20 described later). Based on this, the component name is extracted. The text data of each component name obtained as a result is registered in the component table 113 (FIG. 1 and FIG. 22 described later). This process is the same as the first procedure 1 described above.

  For example, with respect to “Claim 1” in FIG. 7A, as indicated by reference numerals 801 to 802 in FIG. 8, the component element names “reproducing means” “digital voice storing means” “reading means” “voice conversion output means” Is extracted.

Second procedure 2 : Extract the text data describing the component corresponding to each component name extracted in the second procedure 1 from the text data of the claims.
For example, from the text data of “Claim 1” in FIG. 7A, text data of a sentence explaining the component name “reproducing means” is extracted as indicated by 901 to 902 in FIG.

Second procedure 3 : It is determined whether or not sufficient information is included in the text data describing each component, extracted for each component in the claims, extracted in the second procedure 2.
Specifically, for example, with respect to “Claim 1” in FIG. 7A, as shown as 1001 in FIG. 10, the noun group “means” “digital” included in the description of the component “speech conversion means” “Speech”, “analog”, “conversion”, and “output” are extracted by morphological analysis.

  Next, the component name “speech conversion means” itself is subjected to morphological analysis to extract the nouns “speech”, “conversion”, and “means” included in the component name, and these nouns are extracted from the above-described noun 1001. Deleted. As a result, the noun group “digital”, “analog”, and “output” shown as 1002 in FIG. 10 are obtained.

  Next, the noun “digital” which is commonly included in the description of all the constituent elements “reproducing means”, “digital sound storing means”, “reading means”, and “sound conversion output means” in “Claim 1” of FIG. "Is deleted from the noun group obtained as 1002 in FIG. As a result, the noun group “analog” and “output” shown as 1003 in FIG. 10 are obtained.

  In this way, it is determined whether or not the number of nouns included in the noun group remaining as a result of the deletion obtained as 1003 in FIG. 10 is equal to or greater than a predetermined threshold.

Second procedure 4 : For each component, the number of nouns included in the noun group is greater than or equal to a predetermined threshold, and it is determined that the amount of information is sufficiently large and the component is described. The text data in the claims is registered in the component table 113 (FIG. 1 and FIG. 22 described later) in association with the component name. The process combining the second procedure 2 and the second procedure 4 corresponds to the first procedure 2 described above.

Second procedure 5 : If the number of nouns included in the noun group for each component is less than a predetermined threshold, it is determined that the amount of information is small. In this case, the following text data portion is extracted from the document portion of “Example” or “Mode for carrying out the invention” in the patent document extracted in the second procedure 1. That is, the text data portion where the component name is located is extracted before the character string of any expression pattern recorded in the explanatory expression pattern dictionary 115 (FIG. 1 and FIG. 21 described later) in the embodiment. Then, the text data portion following the expression pattern character string is extracted as an explanatory text corresponding to the component name. The expression pattern is, for example, “is”. The text data obtained as a result is registered in the component table 113 (FIG. 1 and FIG. 22 to be described later) as text data explaining the component in association with each component name. In the component element table of FIG. 22, when it is determined that the amount of information is sufficient in the second procedure 4 for the explanation of the component elements, only the explanatory text in the claims described above is registered. Further, when it is determined in the second procedure 5 that the amount of information described in the table is not sufficient, explanations of sections other than the claims described below are also registered.

  For example, when it is determined that the text data of the explanatory text corresponding to the component name “reproduction means” extracted as 902 in FIG. 9 has a small amount of information according to the second procedure 3, for example, the processing illustrated in FIG. Is executed. That is, the text data portion shown as 1102 is extracted from the document portion of the embodiment shown as 1101, and is stored in the component table 113 (FIG. 1 and FIG. 22 described later) as an explanation corresponding to the component name “reproducing means”. It is additionally registered.

Second procedure 6 : Extract text data describing the reason why the effect is obtained from the text data describing the action and effect of the invention in the patent document extracted in the second procedure 1. At this time, by referring to the expression pattern recorded in the reason expression pattern dictionary 116 (FIG. 1 and FIG. 23 described later) from which the effect is obtained, the part of the reason for obtaining the effect is extracted. This process is the same as in the first procedure 3 described above.

  For example, as shown by 1201 to 1202 in FIG. 12, the text data describing the reason why the effect is obtained is extracted for the “effect of the invention” in FIG. 7B. At this time, for example, by using the expression pattern “Koto,” recorded in the reason expression pattern dictionary 116 for obtaining the effect shown in FIG. 23, the text data (1202 in FIG. 12) of the part before the expression pattern is It is extracted as a part where the reason why the effect is obtained is described.

Second procedure 7 : Text data of the reason why the effect extracted in the second procedure 6 is obtained from the text data of the explanatory text of each component registered in the component table 113 in the second procedure 4 or 5. Calculate the most similar. And the component corresponding to the text data of the calculated explanatory note is specified as a component which becomes a point. This process is the same as in the first procedure 4 described above.

  For example, as shown in 1302-1, 1302-2, 1302-3, and 1302-4 in FIG. 13, the “claim 1” in FIG. 7A and the “effect of the invention” in FIG. The text data of the noun is extracted by morphological analysis from the text data of the explanatory text of each component in the claim extracted in the second procedure 4 or 5. On the other hand, as shown as 1301 in FIG. 13, noun text data is extracted by morphological analysis from the text data of the reason why the effect extracted in the second procedure 6 is obtained. The number of nouns included in common with the text data 1301 of the reason for obtaining the effect is similar for each of the text data 1302-1, 1302-2, 1302-3, and 1302-4 of the explanatory text of each component. Counted as degrees. At this time, it can be configured such that, as the frequency of appearance of each noun in the specification of the patent document is lower, the noun is more significant (higher uniqueness) and is weighted higher in the similarity count. As a result, the text data 1302-4 having the largest number of significant nouns included in both the text data of the explanatory text of each component and the text data of the reason for obtaining the effect is extracted as the text data having the highest similarity. The Then, the component name “speech conversion unit” corresponding to the text data 1302-4 is extracted from the component table 113 (FIG. 1 and FIG. 22 described later), and specified as a component serving as a point.

  When the second procedure described above does not include enough information for the similarity determination in the second procedure 4 only by the explanation part of the constituent elements described in the claims, The text data of the explanatory text corresponding to the component is also extracted from the example. As a result, it is possible to extract components that are points with respect to more patents with high accuracy.

  FIG. 14 is an operation flowchart showing a control operation of the embodiment of the component highlighting apparatus having the configuration shown in FIG. The process according to this flowchart is executed by, for example, a computer system equipped with a CPU (Central Processing Unit), a memory, and an external storage device.

  First, the patent database 109 (FIG. 1) is searched to search for a patent document file designated by the user (step S1401). This process realizes the function of the patent document search unit 101 of FIG. The patent database 109 is stored on a storage device (for example, a hard disk storage device). FIG. 18 is a diagram illustrating a data configuration example of the patent database 111. The record data of each patent document registered in the patent database 111 includes, from the application number and the publication number, the summary index, the claim index, the embodiment index, the name data of the invention, and the body data of the patent document. Composed. In addition, the index of an Example may be an index of the means for solving the form for implementing invention, or a subject. The index is, for example, a set of index character strings obtained by combining two adjacent characters from text data corresponding to each part in the patent document based on the bigram algorithm. Users can search patent document files by keyword data such as application number, publication number, words in abstract, words in claims, words in examples, words included in the title of the invention, etc. . When the application number is designated as a search, it is searched whether or not there is a record that matches the character string data of the application number specified as a search in the application number field in the patent database 111. When the public number is designated for retrieval, it is retrieved whether there is a record that matches the character string data of the public number designated for retrieval in the public number field in the patent database 111. When a word included in the abstract, claim, example, or invention name is specified by search, the search specification is made in the abstract, claim, example index field, or invention name field in the patent database 111. It is searched whether the character string data of the entered word is registered. As a result, the text data registered in the body field of the retrieved record is extracted.

  Next, the patent document structure is analyzed for the text data of the patent document extracted in step S1401 (step S1402). This process realizes the function of the patent document structure analysis unit 102 of FIG. For each section such as “claims”, “examples” (“modes for carrying out the invention” or “best mode for carrying out the invention”), “effects of the invention”, patent documents This is a process of analyzing which part of the text data corresponds. Specifically, the text data of each section is extracted from the text data of the patent document as a part from the heading to the next heading by searching for the heading of each section enclosed in black brackets. . A set of sections and text data obtained as a result of the analysis is held on the work memory as an analyzed document table having the data configuration shown in FIG. 19, for example.

  Next, text data corresponding to a claim section (see FIG. 19) among the sections analyzed in step S1402 is acquired from the analyzed document table on the work memory. Next, the component name is extracted from the text data based on the expression pattern recorded in the component expression pattern dictionary 112 (FIG. 1) (step S1403). This process realizes the function of the component name extraction unit 103 in FIG. 1 and corresponds to the first procedure 1 or the second procedure 1 described above. FIG. 20 is a diagram illustrating a data configuration example of the component element expression pattern dictionary 112. In the component element expression pattern dictionary 112, a character string pattern described after (or before) the component element name is recorded as an expression pattern. For example, the word located immediately before this expression pattern character string becomes the component name. Also, for example, the description pattern character string from the immediately preceding to the immediately preceding expression pattern character string (or the head) is the description part of the component corresponding to the component name. Examples of the pattern character string of the component expression pattern dictionary 112 include “having to, to, and”, “having to, to, and”, “to, to, and It is equipped with. The component expression pattern dictionary 112 is held in a work memory or a storage device (such as a hard disk device). In step 1403, pattern matching processing or dependency processing is executed on the text data corresponding to the section of the claim acquired in the work memory by the pattern character string recorded in each entry of the component element pattern dictionary 112. The For example, a noun phrase including a dependency of a noun or a noun located at the end of a text data portion “to” delimited by “to” and a compound word or a modifier is generated, for example, a morpheme. It is extracted as a component name by analysis. The text data of each component name obtained as a result is registered by creating a new entry in the component table 113 (FIG. 1) held in the work memory. FIG. 22 is a diagram illustrating a data configuration example of the component table 113. Each entry of the component table 113 includes a “component name” field, a “component description” field, and a “noun included in the component description” field. In the “component name” field, text data of the component name is registered. In the “description of component” field, text data of an explanation for each component described later is registered. A text data group of nouns obtained by morphological analysis of text data registered in the “description of component” field is registered in the “noun included in the description of component” field. As a specific example, the process of extracting the constituent elements from the claims shown as 801 to 802 in FIG. 8 is executed.

  Next, the component corresponding to each component name extracted in step S1403 from the text data corresponding to the claim section (see FIG. 19) acquired on the work memory in step S1403 will be described. Text data is extracted (step S1404). This process realizes the function of the component description extracting unit 104 from the claims of FIG. 1 and corresponds to the first half process of the first procedure 2 or the second procedure 2 described above. More specifically, in step 1404, a part obtained by deleting the text data of the component name extracted in step S1403 from each text data part “˜” delimited by “to” extracted in step S1403. As described above, the text data portion of the explanatory text is extracted. The extracted text data is held on the work memory.

  Next, the text data of the explanatory text corresponding to the current component name obtained in the work memory in step S1404 is stored in the component name of the component table 113 (FIGS. 1 and 22) on the work memory. It is registered in the “description of component” field of the registration entry (step S1405). This process corresponds to the latter half of the first procedure 2 or the second procedure 4.

  Next, it is determined whether or not sufficient information is included in the text data describing each component in the claim extracted in step S1404 (step S1406). This process realizes the function of the information amount determination unit 105 in FIG. 1 and corresponds to the second procedure 3 described above. FIG. 15 is an operation flowchart showing detailed processing of step S1406.

  In FIG. 15, first, for each component name extracted in step S1403, the text data of the explanation extracted in step S1404 corresponding to each is subjected to morphological analysis, and the morpheme data of the noun in the analysis result is Each component name is held on the work memory (step S1501).

  Next, in the noun group for each component name obtained in step S1501, nouns that are commonly included corresponding to all component names are substituted into the noun string Y, which is a variable in the work memory ( Step S1502).

Next, text data of the first component name among the component names extracted in step S1403 is acquired on the work memory (step S1503).
Next, the text data of the component name acquired in step S1503 or step S1513 described later is subjected to morphological analysis. Then, the noun morpheme data obtained as a result is substituted into the noun string X, which is a variable on the work memory (step S1504).

  Next, the noun morpheme data group included in the explanatory text obtained on the work memory in step S1501 corresponding to the text data of the component name acquired in step S1503 or step S1513 described later is stored on the work memory. Is substituted into a noun string Z that is a variable of (step S1505).

  Next, the morpheme data of each noun in the noun string X on the work memory obtained at step S1504 is deleted from the noun string Z on the work memory obtained at step S1505 (step S1506). This corresponds to the processing examples 1001 to 1002 in FIG. 10 described above, that is, deletion of a noun included in the component name.

  Next, the morpheme data of each noun in the noun string Y on the work memory obtained at step S1502 is deleted from the noun string Z on the work memory obtained at step S1505 (step S1507). This corresponds to the processing examples 1002 to 1003 in FIG. 10 described above, that is, deletion of nouns included in the description of all components.

Subsequently, as a result of step S1507, the number of nouns included in the noun string Z on the work memory is counted (step S1508).
Then, as a result of the counting process in step S1508, it is determined whether or not the number of nouns included in the noun string Z is greater than or equal to a predetermined threshold (step S1509).

  If the number of nouns included in the noun string Z is equal to or greater than a predetermined threshold value, it is determined that there is a sufficient amount of information regarding the component name, and the determination result is associated with the current component name and the work is performed. It is held on the memory (step S1510).

  On the other hand, if the number of nouns included in the noun string Z is less than a predetermined threshold, it is determined that the amount of information regarding the component name is not sufficient, and the determination result is associated with the current component name. Is held on the work memory (step S1511).

  When the information amount determination process corresponding to one component element name is completed by the series of processes of steps S1504 to S1511 described above, it is determined whether or not the process has been completed for all component element names (step S1512).

  If the processing has not been completed for all the component names and the determination in step S1512 is NO, the text data of the next unprocessed component name among the component names extracted in step S1403 is stored in the work memory. Acquired above (step S1513).

Then, the information amount determination process corresponding to the next component element name is repeatedly performed on the component element name again by the series of processes of steps S1504 to S1511.
If the processing is completed for all component names and the determination in step S1512 is YES, the processing in the flowchart in FIG. 15, that is, the processing in step S1406 in FIG.

  Next, the determination result for each component name held in the work memory is confirmed in step S1406 (step S1407). If it is determined that there is a sufficient amount of information for each component, step S1409 is performed. Executed.

  On the other hand, if the determination result that the amount of information is not sufficient is obtained on the work memory in step S1511 of FIG. 15 described above and the determination in step S1407 is NO, the processing in the following step S1408 is performed. After being executed, step S1409 is executed. In step S1408, text data corresponding to the section of the embodiment (see FIG. 19) or the form for carrying out the invention among the sections analyzed in step S1402 is acquired from the analyzed document table on the work memory. Next, the text data portion including the current component name and described by the expression pattern recorded in the explanatory expression pattern dictionary 115 (FIG. 1) in the embodiment is extracted from the text data. The text data obtained as a result is the “descriptive text of the constituent element” field of the registered entry of the constituent element name in the constituent element table 113 (FIGS. 1 and 22) on the work memory as the text data explaining the constituent element. Registered in This process realizes the function of the explanation extracting unit 106 from the embodiment of FIG. 1 and corresponds to the second procedure 5 described above. FIG. 21 is a diagram illustrating a data configuration example of the explanatory expression pattern dictionary 115 in the embodiment. In this dictionary, in the embodiment (see FIG. 19) or in the form for carrying out the invention, a character string of an expression pattern that links a component name and a description corresponding to the component name, for example, “is,” Is recorded. The description is detected assuming that the front of this character string corresponds to the component name and the subsequent corresponds to the description.

  Next, text data describing the reason why the effect is obtained is extracted from the text data corresponding to the section of the effect of the invention (see FIG. 19) or the action of the invention among the sections analyzed in step S1402 (see FIG. 19). Step S1409). This process realizes the function of the reason extraction unit 107 that can obtain the effect of FIG. 1 and corresponds to the first procedure 3 or the second procedure 6 described above. FIG. 16 is a flowchart showing details of the process in step S1409. The process according to this flowchart is executed by, for example, a computer system equipped with a CPU (Central Processing Unit), a memory, and an external storage device.

  In FIG. 16, first, a reason expression pattern for obtaining an effect is read into the work memory from the reason expression pattern dictionary 116 (FIG. 1) for obtaining the effect (step S1601). FIG. 23 is a diagram illustrating a data configuration example of the reason expression pattern dictionary 116 from which an effect is obtained. In this dictionary, in a text part describing the effect of the invention in a patent document, a character string of an expression pattern that separates a character string part describing the effect and a character string part describing the reason for obtaining the effect Is recorded. In the pattern character string field of the reason expression pattern dictionary where the effect is obtained, for example, “so”, “by” and “for” are stored. The part following these expression pattern character strings is a character string part describing the effect. Further, the part before these expression pattern character strings is a character string part describing the reason why the effect is obtained. This dictionary is held in a work memory or a hard disk device. In step 1601, the pattern character string recorded in each entry of the reason expression pattern dictionary 116 for obtaining the effect is read into the work memory.

  Next, text data corresponding to the effect of the invention (see FIG. 19) or the section of the action of the invention in the section analyzed in step S1402 is acquired from the analyzed document table on the work memory (step S1602).

Next, the acquired text data corresponding to the section of the effect or the like of the invention is divided into one or more character strings by the phrase (step S1603).
Next, the first character string among the character strings divided in step S1603 is substituted into the character string r that is a variable on the work memory (step S1604).

  Next, it is determined whether or not the character string r includes any one of the pattern character strings (see FIG. 23) of the reason expression pattern that can obtain any of the effects read in the work memory in step S1601. (Step S1605).

  If the determination in step S1605 is YES, in the character string r, the reason why the effect determined to be included in step S1605 is obtained. The reason why the text data part of the character string before the pattern character string of the expression pattern is effective. Are extracted as text data (step S1606). This text data is held on the work memory. Thereafter, the operation flowchart of FIG. 16, that is, the process of step S1409 of FIG.

If the determination in step S1605 is NO, it is determined whether all the character strings divided in step S1603 have been processed (step S1607).
If all the character strings have not been processed and the determination in step S1607 is NO, the next unprocessed character string among the character strings divided in step S1603 is assigned to the character string r on the work memory ( Step S1608).

Thereafter, a series of processing from step S1605 to S1607 is repeatedly executed.
In the above repetitive processing, if all the character strings divided in step S1603 are processed and the determination in step S1607 is YES, the text data of the reason why the effect is obtained is not extracted, and the flowchart of FIG. In step S1409, the process ends.

  With the above processing, for example, text data describing the reason why the effect is obtained is extracted as shown by 1201 to 1202 in FIG.

  Next, in FIG. 14, if the determination in step S 1407 is YES, in step S 1405, while if the determination in step S 1407 is NO, in step S 1408, each component registered in the component table 113 is displayed. The degree of similarity between the text data of the explanatory text and the text data of the reason for obtaining the effect extracted in step S1409 is calculated (step S1410). This process realizes the function of the similarity calculation unit 108 in FIG. 1 and corresponds to the first half of the first procedure 4 or the second procedure 7. Note that if text data indicating that the effect is obtained in step S1409 is not extracted (when the determination in step S1607 in FIG. 16 is YES), the highlight processing of the components in steps S1410 to S1412 is executed. The process ends without FIG. 17 is a flowchart showing details of the process in step S1410. The process according to this flowchart is executed by, for example, a computer system equipped with a CPU (Central Processing Unit), a memory, and an external storage device.

  First, morpheme analysis is performed on all explanatory texts registered in the “descriptive text of constituent elements” field of the constituent element table 113 illustrated in FIG. The data is registered in the “nouns included in the explanatory text of the component” field of the component table 113 (FIG. 22) (step S1701).

  Next, for the morphological data of each noun calculated in step S1701, the appearance frequency of the patent document read in step S1401 in FIG. 14 in the full text is counted (step S1702).

  Next, the weight of the morpheme data of each noun calculated in step S1701 is calculated as the reciprocal of the appearance frequency counted in step S1702 (step S1703). The weight value obtained as a result is held in a morpheme weight table on the work memory. FIG. 24 is a diagram illustrating a data configuration example of a morpheme weight table. This table can be realized, for example, by two associative array data that outputs the appearance frequency and the weight value using a morpheme character string as a key.

  Next, the text data corresponding to the reason why the effect calculated in step S1409 in FIG. 14 is obtained is subjected to morphological analysis to obtain noun morpheme data, and these are stored in the noun string α held as variables on the work memory. It is substituted (step S1704).

  Next, the text data of the explanatory text registered in the “descriptive text of the constituent element” field of the entry of the first constituent element is read from the constituent element table 113 into the work memory (step S1705).

  Next, the noun group included in the description of the component read in step S1705 is a variable on the work memory from the “noun included in the description of component” field of the corresponding entry on the component table 113. Is read into the noun string β (step S1706).

  Next, common nouns are extracted from the noun string α and the noun string β on the work memory. Then, while referring to and acquiring the morpheme weight table (FIG. 24), the sum of the weight values calculated in step S1703 corresponding to each extracted noun is calculated (step S1707).

  Next, the total value calculated in step S1707 is held in the similarity table 117 on the work memory as the similarity corresponding to the component currently being processed (step S1708). FIG. 25 is a diagram illustrating a data configuration example of the similarity table 117. For each component name, a morpheme character string common to the reason for obtaining the effect and a similarity value are registered.

Thereafter, it is determined whether or not the processing has been completed for all the component entries on the component table 113 (step S1709).
If the determination in step S1709 is NO, the text data of the explanatory text registered in the “descriptive text of the constituent element” field of the entry of the next constituent element is read from the constituent element table 113 into the work memory (step S1710). Then, through a series of processing from step S1706 to S1708, processing for calculating the similarity corresponding to the component is repeatedly executed.

  When the processing is completed for all the component entries on the component table 113 and the determination in step S1709 is YES, the flowchart in FIG. 17, that is, the process in step S1410 in FIG.

  Thereafter, in FIG. 14, the similarity table 117 illustrated in FIG. 25 obtained on the work memory in step S <b> 1410 is referred to, and the component having the highest similarity among the similarities for each component is the point. Is calculated as a constituent element (step S1411). This process realizes the function of the point component specifying unit 109 in FIG. 1 and corresponds to the second half of the first procedure 4 or the second procedure 7.

  Finally, in the patent document read in step S1401, the character string corresponding to the constituent element specified in step S1411 is highlighted and displayed on the display (for example, the output of the input / output device 3003 in FIG. 30). Part). This process realizes the function of the display unit 110 in FIG.

  In the embodiment described above, only nouns are processed as components, but predetermined parts of speech such as verbs, adjectives, adjective verbs, and conjunctions may be used in addition to nouns.

  As the component name, not only a compound word such as “extraction means” but also a phrase such as “extraction means” can be used. In the case of using a phrase, parsing is performed, and a noun phrase is generated by connecting a noun immediately preceding “to” and a modifier related to the noun, and is used as a component name. Further, even if not all modifiers are connected, for example, only one closest modifier may be connected.

  As one of the above-mentioned syntax analysis, the dependency analysis processing is performed by using the flowchart of FIG. 27 and the morpheme dictionary table (a), the part-of-speech connection probability table (b), and the clause connection rule table (c). ) And the phrase connection probability table (d). Each process in the operation flowchart is executed by a computer system including, for example, a CPU, a memory, and an external storage device.

Step S2701 : First, morphological analysis is performed on the constituent elements and their explanations. The morpheme string having the highest probability is selected using a dictionary (a) of morphemes whose elements are parts of speech and character strings such as nouns, particles, and verbs, and a table (b) defining connection probabilities of parts of speech. The morpheme dictionary (a) and the table (b) defining the connection probabilities of parts of speech are stored, for example, on a work memory or a storage device.

Step S2702 : Next, phrase division processing is performed. That is, a phrase is generated by connecting morphemes according to the phrase connection rule (c).

Step S2703 : Next, dependency analysis is performed. That is,
(1) The dependency destination with the highest probability is selected according to the connection probability (d) of the clause and the constraint rule.
(2) The connection probability (d) is defined by the phrase type and proximity relationship.
(3) Perform proximity processing. For example, two morphemes are adjacent, or two morphemes are one segment interval, etc.

On the other hand, a relational element that satisfies the following constraints and has the maximum probability is obtained.
・ A clause other than the last one always has one dependency clause behind. ・ A dependency relationship does not intersect. Finally, a dependency pair between two morphemes is generated.

In the dependency analysis process described above, the dependency pair is generated by pattern matching with the dictionary and calculation of the probability attached to the matched pattern.
By the dependency analysis process described above, for example, a noun phrase is generated by connecting a noun immediately preceding “to” and a modifier related to the noun to form a component name.

  FIG. 29 shows pattern matching processing and characters in steps S1403 and S1404 in FIG. 14, steps S1502, S1506 and S1507 in FIG. 15, steps S1603, S1605 and S1606 in FIG. 16, steps S1702 and S1707 in FIG. It is an operation | movement flowchart which shows the specific operation | movement of a column position detection process. Each process in the operation flowchart is executed by, for example, a computer system including a CPU, a memory, and an external storage device.

  In FIG. 29, first, pattern character string data to be searched is read from a work memory or a storage device into a predetermined area on the work memory (step S2901). For example, in step S1403 of FIG. 14, this pattern character string is a character string of an expression pattern recorded in the component element expression pattern dictionary 112 (FIG. 1) on the storage device.

  Next, data of a matching target character string that is a character string to be subjected to pattern matching processing is read from a work memory or a storage device into a predetermined area on the work memory (step S2902). For example, in step S1403 of FIG. 14, the matching target character string is text data corresponding to a section of a claim acquired from the analyzed document table illustrated in FIG. 14 on the work memory.

  Next, the address of the first character of the pattern character string read in the predetermined area in step S2901 is set in the pattern pointer p1 held in the variable area on the work memory (step 603).

  Next, the address of the first character of the character string to be matched read in the predetermined area in the work memory in step S2902 is set in the matching base point pointer p2 and the matching pointer p3 held in the variable area on the work memory ( Steps S2904, S2905).

  Next, while the address of the matching pointer p3 is incremented by 1 (step S2907), a character that matches the first character in the pattern character string indicated by the pattern pointer p1 is searched from the matching target character string (step S2906). → S2907 → S2908 → S2906 repetitive processing).

  As a result of incrementing the matching pointer p3, if the value of p3 exceeds the end address of the matching target character string read into the predetermined area on the work memory, no matching is output and the processing of FIG. Is finished (step S2908 → S2909).

  When the first character in the pattern character string indicated by the pattern pointer p1 matches in the character string to be matched and the determination in step S2906 becomes YES, the following series of processing from step S2810 to S2817 is executed. Here, it is determined whether or not the character string continuing from the matching position in the matching target character string matches all the second and subsequent characters of the pattern character string.

  First, the address of the matching pointer p3 when a match is detected in step S2906 is set to the matching base point pointer p2 on the work memory (step S2810).

Next, the address value of the pattern pointer p1 on the work memory is incremented by 1 (step S2811).
Next, it is determined whether or not the address value of the pattern pointer p1 has exceeded the last address of the pattern character string read into the predetermined area on the work memory (step S2812).

If the determination in step S2812 is NO, the address value of the matching pointer p3 is incremented by 1 (step S2813).
Next, it is determined whether or not the address value of the matching pointer p3 exceeds the end address of the matching target character string read into the predetermined area on the work memory (step S2815).

  If the determination in step S2814 is NO, it is determined whether or not the character at the address position in the matching target character string pointed to by the matching pointer p3 matches the character at the address position in the pattern character string pointed to by the pattern pointer p1. (Step S2817). That is, when step S2817 is executed first after step S2906 becomes YES, the second character counted from the position where the matching position is detected in step S2906 in the matching target character string is the two characters of the pattern character string. It is determined whether or not the eye matches.

  If the determination in step S2817 is YES, the process returns to step S2811, and a series of processing up to step S2817 is executed again. Here, the address values of the pattern pointer p1 and the matching pointer p3 are incremented by 1 (steps S2811, S2814). As a result, it is further determined whether or not the third character counted from the position where the matching position is detected in step S2906 in the matching target character string matches the third character of the pattern character string.

  By repeating the above processing, it is determined whether or not the character string continuing from the position where the match is detected in step S2906 in the matching target character string matches all the second and subsequent characters of the pattern character string.

  In the above series of processing, if the value of the pattern pointer p1 incremented in step S2811 exceeds the end address of the pattern character string read in the predetermined area on the work memory, the pattern character on the matching target character string A match with the column is detected. In this case, the address value of the matching base point pointer p2 indicating the position where a match is detected in step S2906 in the matching target character string is output as a matching result, and the processing in FIG. 29 is terminated (steps S2812 → S2813).

  On the other hand, if the value of the matching pointer p3 incremented in step S2814 exceeds the end address of the matching target character string read into the predetermined area on the work memory, no matching is output and the processing in FIG. Is terminated (steps S2815 → S2816).

  In the above-described series of processing, the character at the address position in the matching target character string pointed to by the matching pointer p3 does not match the character at the address position in the pattern character string pointed to by the pattern pointer p1, and the determination in step S2817 is NO. If it becomes, re-search is executed. That is, the address of the first character of the pattern character string read into the predetermined area on the work memory in step S2901 is set to the pattern pointer p1 (step 618). Further, the address value corresponding to the position next to the position where the match is detected in step S2906 pointed to by the matching base point pointer p2 is set in the matching pointer p3 (step S2819). Thereafter, the processing returns to step S2906, and the following processing is executed while the address of the matching pointer p3 is incremented by 1 from the position next to the matching position (step S2907). That is, the character that matches the first character in the pattern character string indicated by the pattern pointer p1 is searched again from the character string to be matched (steps S2906 → S2907 → S2908 → S2906 repetitive processing).

  When the first character in the pattern character string indicated by the pattern pointer p1 matches again in the matching target character string and the determination in step S2906 becomes YES, the following series of processing from step S2810 to S2817 is executed. Thereby, it is determined again whether or not the character string continuing from the matching position in the matching target character string matches all the second and subsequent characters of the pattern character string.

  By repeating the above-described series of processing, a pattern character string is searched for in the matching target character string. If matching is successful, the head position of the match is output as a matching result (step S2813).

FIG. 30 is a diagram illustrating an example of a hardware configuration of a computer that can implement the system according to each of the above-described embodiments.
The computer shown in FIG. 30 includes a CPU 3001, a memory 3002, an input / output device 3003, an external storage device 3005, a portable recording medium driving device 3006 into which a portable recording medium 3009 is inserted, and a communication interface 3007. The buses 3008 are connected to each other.

  The CPU 3001 controls the entire computer. The memory 3002 has a work area for temporarily storing a program or data stored in the external storage device 3005 (or portable recording medium 3009) when executing a program, updating data, or the like. The CUP 3001 performs overall control by reading the program into the memory 3002 and executing it.

  The input / output device 3003 detects an input operation by a user using a keyboard, a mouse, or the like, notifies the CPU 3001 of the detection result, and outputs data sent under the control of the CPU 3001 to a display device or a printing device.

The external storage device 3005 is, for example, a hard disk storage device. Mainly used for storing various data and programs.
The portable recording medium driving device 3006 accommodates a portable recording medium 3009 such as an optical disk, SDRAM, or Compact Flash (registered trademark), and has an auxiliary role for the external storage device 3005.

The communication interface 3007 is a device for connecting, for example, a LAN (local area network) or WAN (wide area network) communication line.
The system according to the present embodiment is realized by the CPU 3001 executing a program corresponding to the operation flowchart for realizing the functions of the embodiment. The program may be recorded and distributed in, for example, the external storage device 3005 or the portable recording medium 3009, or may be acquired from the network by the communication interface 3007.

Regarding the above embodiment, the following additional notes are disclosed.
(Appendix 1)
A device for supporting understanding of patent documents,
The text data of the claims in the patent document extracted from the patent document database using the component expression pattern dictionary that records the character string pattern described before or after the component name corresponding to the component constituting the claim A component name extraction unit that extracts each component name from
Description that explains the component corresponding to each component name from the text data of the claim based on each component name and the character string pattern recorded in the component expression pattern dictionary A component description extractor from the claim for extracting sentence text data;
Using a reason expression pattern dictionary that provides an effect of recording a character string pattern that links a character string portion that describes an effect and a character string portion that describes the reason for obtaining the effect. A reason extraction unit capable of obtaining the effect of extracting the text data describing the reason for obtaining the effect from the text data describing the effect of the invention;
A similarity calculation unit that calculates the similarity between the text data of the description of each component and the text data describing the reason why the effect is obtained;
A point component identifying unit that identifies a component corresponding to the text data of the explanatory text having the highest similarity as a component serving as a point;
An output unit for outputting the constituent elements serving as the points;
A component highlighting device comprising:
(Appendix 2)
A device for supporting understanding of patent documents,
The text data of the claims in the patent document extracted from the patent document database using the component expression pattern dictionary that records the character string pattern described before or after the component name corresponding to the component constituting the claim A component name extraction unit that extracts each component name from
Description that explains the component corresponding to each component name from the text data of the claim based on each component name and the character string pattern recorded in the component expression pattern dictionary A component description extractor from the claim for extracting sentence text data;
An information amount determination unit for determining whether or not sufficient information is included in the text data of the explanatory text describing each component;
In the detailed description of the invention in the patent document, the description expression pattern dictionary in the embodiment in which the character string pattern that links the component name and the description corresponding to the component name is recorded is used. Explanation from an embodiment in which text data of an explanatory text corresponding to a component name determined that sufficient information is not included in the text data of the explanatory text is extracted from text data describing a detailed explanation An extractor;
Using the reason expression pattern dictionary that provides the effect of recording the character string pattern that links the character string part describing the effect and the character string part describing the reason for obtaining the effect, the invention in the patent document A reason extraction unit that obtains the effect of extracting the text data describing the reason for obtaining the effect from the text data describing the effect;
A similarity calculation unit that calculates the similarity between the text data of the description of each component and the text data describing the reason why the effect is obtained;
A point component identifying unit that identifies a component corresponding to the text data of the explanatory text having the highest similarity as a component serving as a point;
An output unit for outputting the constituent elements serving as the points;
A component highlighting device comprising:
(Appendix 3)
The information amount determination unit corresponds to the component from the morphological data of a predetermined part of speech obtained by morphological analysis of the text data of the explanatory text explaining the component for each component. The number of remaining morpheme data of the predetermined part-of-speech obtained by deleting the morpheme data included in the component name and the morpheme data commonly included in the explanations of all other component elements is equal to or greater than a predetermined threshold. By determining whether or not there is sufficient information included in the text data of the explanatory text describing each component,
The component highlighting device according to appendix 2, characterized in that:
(Appendix 4)
The similarity calculation unit
Extracting morpheme data of a predetermined part of speech by morphological analysis from the text data of the description corresponding to each component name,
From text data describing the reason why the effect is obtained, morphological data of a predetermined part of speech is extracted by morphological analysis,
The number of morpheme data of the predetermined part-of-speech extracted in common between the morpheme data in the explanatory text corresponding to each component name and the morpheme data describing the reason why the effect is obtained. Calculating the similarity between the text data of the explanatory text corresponding to the name and the text data describing the reason why the effect is obtained;
3. The component highlighting device according to appendix 1 or 2, wherein
(Appendix 5)
The similarity calculation unit
For each of the morpheme data of the predetermined part of speech, a weight that increases as the frequency of appearance in the patent document decreases, and the number of morpheme data of the predetermined part of speech extracted in common is set for each morpheme data. Calculate by adding the weight of
The component highlighting device according to appendix 6, wherein:
(Appendix 6)
To support the understanding of patent documents,
Extracted from a patent document database using a component element expression pattern dictionary in which a character string of an expression pattern in which a character string pattern described before or after a component name corresponding to a component element constituting a claim is recorded is recorded Each component name is extracted from the text data of the claims in the patent document
Description that explains the component corresponding to each component name from the text data of the claim based on each component name and the character string pattern recorded in the component expression pattern dictionary Extract text data of sentence,
Using a reason expression pattern dictionary that provides an effect of recording a character string pattern that links a character string portion that describes an effect and a character string portion that describes the reason for obtaining the effect. Extracting the text data describing the reason why the effect is obtained from the text data describing the effect of the invention,
Calculating the similarity between the text data of the description of each component and the text data describing the reason why the effect is obtained;
A component corresponding to the text data of the explanatory note having the highest similarity is identified as a component that becomes a point,
Output the component that is the point;
A program for executing functions.
(Appendix 7)
To support the understanding of patent documents,
The text data of the claims in the patent document extracted from the patent document database using the component expression pattern dictionary that records the character string pattern described before or after the component name corresponding to the component constituting the claim Extract each component name from
Description that explains the component corresponding to each component name from the text data of the claim based on each component name and the character string pattern recorded in the component expression pattern dictionary Extract text data of sentence,
It is determined whether or not sufficient information is included in the text data of the explanatory text describing each component,
In the detailed description of the invention in the patent document, the description expression pattern dictionary in the embodiment in which the character string pattern that links the component name and the description corresponding to the component name is recorded is used. Extracting the text data of the explanatory text corresponding to the component name determined that the text data of the explanatory text does not contain sufficient information from the text data describing the detailed explanation,
Using the reason expression pattern dictionary that provides the effect of recording the character string pattern that links the character string part describing the effect and the character string part describing the reason for obtaining the effect, the invention in the patent document Extract text data describing the reason for obtaining the effect from text data describing the effect,
Calculating the similarity between the text data of the description of each component and the text data describing the reason why the effect is obtained;
A component corresponding to the text data of the explanatory note having the highest similarity is identified as a component that becomes a point,
Output the component that is the point;
A program for executing functions.
(Appendix 8)
In the determination of the amount of information, for each component, the morpheme data of a predetermined part-of-speech obtained by morphological analysis of the text data of the explanatory text explaining the component, the corresponding to the component The number of remaining morpheme data of the predetermined part-of-speech obtained by deleting the morpheme data included in the component name and the morpheme data commonly included in the explanations of all other component elements is equal to or greater than a predetermined threshold. By determining whether or not there is sufficient information included in the text data of the explanatory text describing each component,
The program according to appendix 7, characterized by:
(Appendix 9)
In the calculation of the similarity,
Extracting morpheme data of a predetermined part of speech by morphological analysis from the text data of the description corresponding to each component name,
From text data describing the reason why the effect is obtained, morphological data of a predetermined part of speech is extracted by morphological analysis,
The number of morpheme data of the predetermined part-of-speech extracted in common between the morpheme data in the explanatory text corresponding to each component name and the morpheme data describing the reason why the effect is obtained. Calculating the similarity between the text data of the explanatory text corresponding to the name and the text data describing the reason why the effect is obtained;
The program according to appendix 6 or 7, characterized by:
(Appendix 10)
In the calculation of the similarity,
For each of the morpheme data of the predetermined part of speech, a weight that increases as the frequency of appearance in the patent document decreases, and the number of morpheme data of the predetermined part of speech extracted in common is set for each morpheme data. Calculate by adding the weight of
The program according to appendix 9, characterized by:
(Appendix 11)
A method for a computer to assist in understanding patent documents,
Extracted from a patent document database using a component element expression pattern dictionary in which a character string of an expression pattern in which a character string pattern described before or after a component name corresponding to a component element constituting a claim is recorded is recorded Each component name is extracted from the text data of the claims in the patent document
Description that explains the component corresponding to each component name from the text data of the claim based on each component name and the character string pattern recorded in the component expression pattern dictionary Extract text data of sentence,
Using a reason expression pattern dictionary that provides an effect of recording a character string pattern that links a character string portion that describes an effect and a character string portion that describes the reason for obtaining the effect. Extracting the text data describing the reason why the effect is obtained from the text data describing the effect of the invention,
Calculating the similarity between the text data of the description of each component and the text data describing the reason why the effect is obtained;
A component corresponding to the text data of the explanatory note having the highest similarity is identified as a component that becomes a point,
Output the component that is the point;
A component highlighting method, wherein the computer executes the above.
(Appendix 12)
A method for a computer to assist in understanding patent documents,
The text data of the claims in the patent document extracted from the patent document database using the component expression pattern dictionary that records the character string pattern described before or after the component name corresponding to the component constituting the claim Extract each component name from
Description that explains the component corresponding to each component name from the text data of the claim based on each component name and the character string pattern recorded in the component expression pattern dictionary Extract text data of sentence,
It is determined whether or not sufficient information is included in the text data of the explanatory text describing each component,
In the detailed description of the invention in the patent document, the description expression pattern dictionary in the embodiment in which the character string pattern that links the component name and the description corresponding to the component name is recorded is used. Extracting the text data of the explanatory text corresponding to the component name determined that the text data of the explanatory text does not contain sufficient information from the text data describing the detailed explanation,
Using the reason expression pattern dictionary that provides the effect of recording the character string pattern that links the character string part describing the effect and the character string part describing the reason for obtaining the effect, the invention in the patent document Extract text data describing the reason for obtaining the effect from text data describing the effect,
Calculating the similarity between the text data of the description of each component and the text data describing the reason why the effect is obtained;
A component corresponding to the text data of the explanatory note having the highest similarity is identified as a component that becomes a point,
Output the component that is the point;
A component highlighting method, wherein the computer executes the above.
(Appendix 13)
In the determination of the amount of information, for each component, the morpheme data of a predetermined part-of-speech obtained by morphological analysis of the text data of the explanatory text explaining the component, the corresponding to the component The number of remaining morpheme data of the predetermined part-of-speech obtained by deleting the morpheme data included in the component name and the morpheme data commonly included in the explanations of all other component elements is equal to or greater than a predetermined threshold. By determining whether or not there is sufficient information included in the text data of the explanatory text describing each component,
Item 14. The component highlighting method according to Appendix 12.
(Appendix 14)
In the calculation of the similarity,
Extracting morpheme data of a predetermined part of speech by morphological analysis from the text data of the description corresponding to each component name,
From text data describing the reason why the effect is obtained, morphological data of a predetermined part of speech is extracted by morphological analysis,
The number of morpheme data of the predetermined part-of-speech extracted in common between the morpheme data in the explanatory text corresponding to each component name and the morpheme data describing the reason why the effect is obtained. Calculating the similarity between the text data of the explanatory text corresponding to the name and the text data describing the reason why the effect is obtained;
The component highlighting method according to appendix 11 or 12, wherein the computer executes the above.
(Appendix 15)
In the calculation of the similarity,
For each of the morpheme data of the predetermined part of speech, a weight that increases as the frequency of appearance in the patent document decreases, and the number of morpheme data of the predetermined part of speech extracted in common is set for each morpheme data. Calculate by adding the weight of
The component highlighting method according to claim 14, wherein the computer executes the above.

DESCRIPTION OF SYMBOLS 101 Patent document search part 102 Patent document structure analysis part 103 Constituent element name extraction part 104 Description extraction part of the constituent element from Claim 105 Information amount determination part 106 Explanation extraction part from Example 107 Reason extraction part 108 from which an effect is acquired Similarity calculation unit 1
DESCRIPTION OF SYMBOLS 109 Point component specific part 110 Display part 111 Patent database 112 Component element expression pattern dictionary 113 Component element table 114 Analyzed document table 115 Explanation expression pattern dictionary 116 Example effect expression pattern dictionary 117 effect degree table Similarity table

Claims (6)

A device for supporting understanding of patent documents,
The text data of the claims in the patent document extracted from the patent document database using the component expression pattern dictionary that records the character string pattern described before or after the component name corresponding to the component constituting the claim A component name extraction unit that extracts each component name including a noun phrase connecting a noun and one or more modifiers associated with the noun and the closest modifier.
Description that explains the component corresponding to each component name from the text data of the claim based on each component name and the character string pattern recorded in the component expression pattern dictionary A component description extractor from the claim for extracting sentence text data;
Using a reason expression pattern dictionary that provides an effect of recording a character string pattern that links a character string portion that describes an effect and a character string portion that describes the reason for obtaining the effect. A reason extraction unit capable of obtaining the effect of extracting the text data describing the reason for obtaining the effect from the text data describing the effect of the invention;
A similarity calculation unit that calculates the similarity between the text data of the description of each component and the text data describing the reason why the effect is obtained;
A point component identifying unit that identifies a component corresponding to the text data of the explanatory text having the highest similarity as a component serving as a point;
An output unit for outputting the constituent elements serving as the points;
A component highlighting device comprising: A device for supporting understanding of patent documents,
The text data of the claims in the patent document extracted from the patent document database using the component expression pattern dictionary that records the character string pattern described before or after the component name corresponding to the component constituting the claim A component name extraction unit that extracts each component name from
Description that explains the component corresponding to each component name from the text data of the claim based on each component name and the character string pattern recorded in the component expression pattern dictionary A component description extractor from the claim for extracting sentence text data;
For each component, the morpheme included in the component name corresponding to the component from the morpheme data of a predetermined part of speech obtained by morphological analysis of the text data of the explanatory text describing the component Determining whether the number of remaining morpheme data of the predetermined part-of-speech obtained by deleting data and morpheme data commonly included in the explanations of all other components is equal to or greater than a predetermined threshold An information amount determination unit that determines whether or not sufficient information is included in the text data of the explanatory text describing each component,
In the detailed description of the invention in the patent document, the description expression pattern dictionary in the embodiment in which the character string pattern that links the component name and the description corresponding to the component name is recorded is used. Explanation from an embodiment in which text data of an explanatory text corresponding to a component name determined that sufficient information is not included in the text data of the explanatory text is extracted from text data describing a detailed explanation An extractor;
Using the reason expression pattern dictionary that provides the effect of recording the character string pattern that links the character string part describing the effect and the character string part describing the reason for obtaining the effect, the invention in the patent document A reason extraction unit that obtains the effect of extracting the text data describing the reason for obtaining the effect from the text data describing the effect;
Similarity between the text data of the explanatory text of each constituent element extracted by the explanation extracting section of the constituent element from the claim or the explanation extracting section from the embodiment and the text data describing the reason why the effect is obtained A similarity calculator for calculating the degree;
A point component identifying unit that identifies a component corresponding to the text data of the explanatory text having the highest similarity as a component serving as a point;
An output unit for outputting the constituent elements serving as the points;
A component highlighting device comprising: The similarity calculation unit
Extracting morpheme data of a predetermined part of speech by morphological analysis from the text data of the description corresponding to each component name,
From text data describing the reason why the effect is obtained, morphological data of a predetermined part of speech is extracted by morphological analysis,
The number of morpheme data of the predetermined part-of-speech extracted in common between the morpheme data in the explanatory text corresponding to each component name and the morpheme data describing the reason why the effect is obtained. Calculating the similarity between the text data of the explanatory text corresponding to the name and the text data describing the reason why the effect is obtained;
The component highlighting device according to claim 1, wherein the component highlighting device is a component highlighting device. The similarity calculation unit
For each of the morpheme data of the predetermined part of speech, a weight that increases as the frequency of appearance in the patent document decreases, and the number of morpheme data of the predetermined part of speech extracted in common is set for each morpheme data. Calculate by adding the weight of
The component highlighting device according to claim 3. To support the understanding of patent documents,
Extracted from a patent document database using a component element expression pattern dictionary in which a character string of an expression pattern in which a character string pattern described before or after a component name corresponding to a component element constituting a claim is recorded is recorded Each component name including a noun phrase connecting a noun and one or more modifiers closest to the noun from the text data of the claims in the patent document ,
Description that explains the component corresponding to each component name from the text data of the claim based on each component name and the character string pattern recorded in the component expression pattern dictionary Extract text data of sentence,
Using a reason expression pattern dictionary that provides an effect of recording a character string pattern that links a character string portion that describes an effect and a character string portion that describes the reason for obtaining the effect. Extracting the text data describing the reason why the effect is obtained from the text data describing the effect of the invention,
Calculating the similarity between the text data of the description of each component and the text data describing the reason why the effect is obtained;
A component corresponding to the text data of the explanatory note having the highest similarity is identified as a component that becomes a point,
Output the component that is the point;
A program for executing functions. A method for a computer to assist in understanding patent documents,
Extracted from a patent document database using a component element expression pattern dictionary in which a character string of an expression pattern in which a character string pattern described before or after a component name corresponding to a component element constituting a claim is recorded is recorded Each component name including a noun phrase connecting a noun and one or more modifiers closest to the noun from the text data of the claims in the patent document ,
Description that explains the component corresponding to each component name from the text data of the claim based on each component name and the character string pattern recorded in the component expression pattern dictionary Extract text data of sentence,
Using a reason expression pattern dictionary that provides an effect of recording a character string pattern that links a character string portion that describes an effect and a character string portion that describes the reason for obtaining the effect. Extracting the text data describing the reason why the effect is obtained from the text data describing the effect of the invention,
Calculating the similarity between the text data of the description of each component and the text data describing the reason why the effect is obtained;
A component corresponding to the text data of the explanatory note having the highest similarity is identified as a component that becomes a point,
Output the component that is the point;
A component highlighting method, wherein the computer executes the above.